An evaluation of structured navigation for subject searching in online catalogues
Understanding and improving subject searching in online library catalogues is the focus of this study. Against the backdrop of current research and developments in online catalogues an analysis of the problems and prospects for subject access in the expanding online catalogue is presented. Developments in recent information retrieval theory and practice are reviewed, and a case is made for a new model of information seeking and retrieval that more closely describes much of the subject searching and browsing activity actually conducted by library users. The center piece of this study is the experiment that was conducted using an experimental online catalogue developed to investigate and evaluate the effect of alternative browse and navigate search methods on overall retrieval effectiveness and subject searching performance. The objectives, methodology, and findings of this online catalogue search experiment are discussed. The primary aim of the experimental study was to evaluate the usability and retrieval performance of a pre-structured "navigation" approach to subject searching and browsing in library catalogues. The main hypothesis tested was that the provision and use of a navigation search and browse function would significantly improve overall OPAC retrieval effectiveness and the subject searching performance of OPAC users. The OPAC used in the study was designed and implemented by this author using the database management and retrieval software known as "TiNMAN", provided by Information Management & Engineering, Ltd. TINMAN employs an entity-relational database structure which permits the linking of any field in the stored bibliographic record to any other field. These linkages establish browse and navigation pathways among data fields ("entities") and citations to support guided but flexible searching and browsing through the collection by users. Thus, a rudimentary form of hypertext is provided for the users of the OPAC. The test database consisted of 30,000 Library of Congress MARC bibliographic records selected at random from all LC catalog records for publications through 1988 in the English language in the LC classes HB-HJ (Economics, Business, etc.). For each record, the verbal description of the assigned LC class number found in the printed schedules was added as a subject descriptor to augment the subject cataloging provided by the Library of Congress. Three different OPACs were tested for comparison purposes. The control OPAC lacked the navigation feature. The other two OPACs supported related-record navigation, one on title words only, the other on subject headings only. Searchers were encouraged to use the OPAC's features and search options in whatever manner they wished. Subjects in Group-I were permitted to navigate only on the subject headings from the controlled subject vocabulary assigned to the work cited (augmented by the verbal meanings of the Library of Congress class number). Subjects in Group-2 were permitted to navigate, but only from title words of the work cited and displayed. Navigating from one of these title words would result in the retrieval of all works whose titles had at least one occurrence of the selected word. Subjects in the control group were not permitted to navigate; that is, it was not possible for them to point to a selected data element in a displayed citation to move on to related terms or citations associated with that data element. The positive value of related-record navigation in improving subject searching in OPACs was not clearly determined. The navigation groups performed significantly better than the control groupon the first search task, but all three groups performed nearly equally well on the second search task. Navigation on subject headings or title keywords resulted in higher recall scores, especially among first time, novice users of the system, but precision suffered significantly in title-word navigation. In fact, the control group achieved higher precision scores on both search tasks. Navigation did not seem to aid subject searching performance after greater familiarity with the system was achieved, except perhaps to increase recall in persistent searches without much decrease in precision. Online bookshelf browsing seems to improve recall without a significant decrease in precision, and may be a more positive factor than navigation on either subject headings or title words.